Families of symmetric periodic orbits in the three body problem and

Anuncio
Monografı́as de la Real Academia de Ciencias de Zaragoza. 25: 229–240, (2004).
Families of symmetric periodic orbits in the three body problem
and the figure eight
F. J. Muñoz-Almaraz† , J. Galán‡ and E. Freire‡
† Departamento de Matemática Aplicada y Computación.
Facultad de Ciencias. Universidad de Valladolid.
47005 Valladolid, Spain.
‡ Departamento de Matemática Aplicada II.
Escuela Superior de Ingenieros. Universidad de Sevilla.
41092 Sevilla, Spain.
Abstract
In this paper we show a technique for the continuation of symmetric periodic orbits in systems with time-reversal symmetries. The geometric idea of this technique
allows us to generalize the “cylinder” theorem for this kind of systems. We state
the main theoretical result without proof (to be published elsewhere). We focus on
the application of this scheme to the three body problem (TBP), taking as starting
point the figure eight orbit [3] to find families of symmetric periodic orbits.
Key words and expressions: Hamiltonian and conservative systems, periodic
solutions, numerical continuation, boundary value problems, three-body problem,
figure-8 orbit.
MSC: 34C25, 34C30, 34C14, 37J15, 65L10.
1
Introduction
Our interest is about finding symmetric periodic solutions in the three body problem. In
a previous paper [7] we studied the continuation of periodic orbits in the TBP following
the scheme given in [13] for continuation in conservative systems. Unlike other methods,
this scheme does not make use of symplectic reduction before numeric calculation. Several
families of periodic orbits were shown in [7] and the numerical study of their linear stability
(characteristic multipliers). Another applications of this technique appeared in [6] where
229
a lot of families of periodic orbits in the TBP and the restricted three body problem
were found. Some of these families are formed by symmetric periodic orbits with respect
to time-reversal symmetries. However, while the scheme given in [13] does not take
advantage of these symmetries, classical methods do [1, 4]. Our scope in this paper is to
adapt the scheme in [13], so that families of symmetric periodic orbits can be computed
in a more proper way than in [7, 6] where time-reversal symmetries were not used for the
computations.
Chenciner and Montgomery [3] proved recently the existence of a spectacular solution
in the TBP called the figure eight, since the three bodies follow the same curve (choreography) with this shape. The initial conditions of the figure eight, which were computed
numerically by Simó, are shown in [3] too. This kind of solution had been predicted by
Moore [11] in 1993. A lot of work has been developed around this new solution for the
last two years. Simó also studied the region of stability near the figure eight and found
countless examples of choreographies [14, 15]. For other results on continuation, see [9, 2].
The figure eight is symmetric with respect to several time-reversal symmetries; therefore, we think that it is a good starting point to look for symmetric periodic solutions.
In this introduction we give a short revision about the main concepts related to symmetries in the TBP and we describe with respect to which time-reversal symmetries the
figure eight is symmetric (for more details, see [3, 9]).
In the second section we show geometrically when a first integral plays a role in the
continuation of symmetric periodic orbits. In the third section we are going to state the
main theoretical result in this paper about persistence (or continuation) of symmetric
periodic orbits. As we have said in the abstract, this result is a generalization of the
“cylinder” theorem. We adapt the ideas from reference [13] where continuation of periodic
orbits in conservative systems is based on a proper unfolding of these systems. For
symmetric periodic orbits continuation it is necessary to unfold systems with first integrals
which verify a certain property. In the last section we apply these ideas to the figure eight
in the TBP.
1.1
On symmetries in the TBP
Let n be the dimension of the configuration space, so if n = 2, the planar TBP is set and
if n = 3, it is the spatial TBP. For each smooth function F : R6n 7→ R the vector field
XF (x) = J∇F (x) is defined, where J is the matrix of size 6n × 6n

J =
03n −I3n
I3n
230
03n

.
(1)
The TBP (for bodies of masses mj with j = 1, 2, 3) is a system ẋ = XH (x) whose
Hamiltonian H is given by
H(x) =
3
X
X
1
mi mj
kpj k2 −
,
j=1 2mj
1≤i<j≤3 kqi − qj k
(2)
with x = (q1 , q2 , q3 , p1 , p2 , p3 ) where qj and pj are the position vector and momentum
vector of the j-th body (with j = 1, 2, 3).
Besides the Hamiltonian, the TBP has other first integrals, as the components of the
total linear momentum and the total angular momentum. Moreover, every quadratic first
integral can be written as

Fa,A (x) = a∗ 
3
X
j=1

pj  +
3
X
q∗j Apj ,
(3)
j=1
where a ∈ Rn and A is a skew symmetric matrix. Moreover, in this paper, the transpose
of a vector x ∈ Rn is denoted by x∗ . The vectors of the canonical basis in R2 are denoted
as e1 and e2 . In a similar way, for R3 are denoted as e1 , e2 and e3 .
The Noether’s symmetry theorem [10] states that the flux of ẋ = XFa,A (x) is a symplectic continuous group of symmetries, i.e. the system ẋ = XH (x) remains invariant
under the changes of variables induced from the flux of ẋ = XFa,A (x).
We can easily check that ΨQ,b are symmetries of the TBP
ΨQ,b : R6n →
Rn ×
Rn
×
Rn
× Rn × Rn ×Rn
x 7→ (Qq1 + b , Qq2 + b , Qq3 + b , Qp1 , Qp2 , Qp3 )
(4)
where Q is an orthogonal matrix of size n × n and b is any vector in Rn .
If two bodies have equal masses, then there is another symmetry which exchanges
positions and momenta for the bodies. For instance, with m2 = m3 , the TBP has the
symmetry
C : (q1 , q2 , q3 , p1 , p2 , p3 ) 7→ (q1 , q3 , q2 , p1 , p3 , p2 ).
(5)
We recall that the composition of symmetries is a symmetry.
A matrix R of size 6n×6n such that R2 = I6n is said to be a time-reversal symmetry of
the TBP iff XH (R x) = −R XH (x). The fixed point subspace of a time-reversal symmetry
is Fix(R) = {x ∈ R6n : Rx = x}. In the TBP a time-reversal symmetry is

RN = 
I3n
03n
03n −I3n

.
(6)
Any composition of a symmetry with RN is another time-reversal symmetry. Therefore,
the TBP is full of time-reversal symmetries. It is well known (see [8]) that an orbit with
two points in Fix(R) is a symmetric periodic orbit, i.e. the orbit is invariant with respect
to R.
231
In the TBP there exists a rescale Φλ (with λ ∈ R \ {0}) of the positions and the
conjugate momenta which maps a solution into another solution with a new time scale.
This function is
R6n
Φλ :
R6n
→
(q1 , q2 , q3 , p1 , p2 , p3 ) 7→ (λ−2 q1 , λ−2 q2 , λ−2 q3 , λp1 , λp2 , λp3 ),
(7)
and its time–rescale is λ−3 .
1.2
The figure eight as a symmetric periodic orbit
Chenciner and Montgomery [3] looked for a planar curve which is a minimum of the action
for the TBP with the bodies of unit mass,
Z
0
T
(
3
X
1
1X
) dt,
kpj (t)k2 +
2 j=1
1≤i<j≤3 kqi (t) − qj (t)k
(8)
over the subspace of curves which verify that:
• Initial condition: The first body is in the middle of the other two bodies.
• Final condition: The second body (which was in an extremum in the initial condition) is now equidistant of the first and the third bodies.
They proved that the minimum is reached in a curve which points generate a solution
of the TBP where the three bodies follow the same curve (choreography) with a shape of
an eight, so this orbit was called the figure eight by them. In Figure 1, we see the curve
followed by the bodies and the initial and final conditions for the minimization problem.
0.4
0.3
0.4
m
1
0.2
0.1
0.1
m
0
0
1
−0.1
m
3
−0.1
−0.2
−0.2
−0.3
−0.4
−1.5
m
0.3
3
0.2
m
−0.3
1
−0.4
−1.5
2
−1
−0.5
0
0.5
1.5
Initial condition
m
2
−1
−0.5
0
0.5
1
1.5
Final condition
Figure 1: Figure eight with the initial and final conditions of the minimization problem.
The figure eight is symmetric with respect to six time-reversal symmetries in the planar
case and nine time-reversal symmetries in the spatial case. These time-reversal symmetries
are given by the composition of RN with an exchange of the bodies (for example, C is one
232
of the three possible exchanges) and with a symmetry ΨS,0 . The matrix S can be chosen
from the following list for the planar case:

SM1 = 
1
0
0 −1

 , SE
1

=
−1
0
0
−1

(9)

and for the spatial case:

1
0

SM1 3D = 
 0 −1

0
0
0


1 0
0


−1







, SΠ1 3D = 
 0 1 0  and SE1 3D =  0
0 




−1
0 0 −1
0
0
−1
0
0



. (10)
0 

−1
In order to simplify (and following [3]), the time-reversal symmetry RN ◦ C ◦ ΨSM1 ,0
is called M1 and RN ◦ C ◦ ΨSE1 ,0 is called E1 throughout this paper. The fixed points of
M1 have the first body in the axis e1 and this body forms an isosceles triangle with the
other two bodies whose symmetry axis is e1 ; the conjugate momentum of the position of
the first body, p1 , is in the direction e2 and the conjugate momenta of the three bodies
(p1 , p2 and p3 ) form an isosceles triangle with symmetry axis in e2 . See Figure 2 for a
graphical interpretation.
p2
p3
p1
p2
p1
e2
e1
p3
Figure 2: Fixed points of M1 .
The fixed points of E1 are given by the first body in the origin and in the middle
point of the other two bodies; the conjugate momenta of this two bodies are identical,
i.e., q1 = 0, q2 = −q3 and p2 = p3 .
2
Geometric Interpretation of the Continuation with Time-Reversal Symmetries
There exists a “cylinder” theorem for systems with a time-reversal symmetry (see [8]). A
simple geometric reasoning states that periodic orbits generically arise as a one-parameter
family (we are assuming that the dimension of Fix(R) is half the whole dimension).
233
However, the TBP is a non-generic system, since most of the orbits do not belong to oneparameter families. We have studied why some first integrals avoid a generic behavior for
the TBP.
We focus on the time-reversal symmetry M1 for our explanation. An analogous situation appears for E1 . A translation of a fixed point of M1 in the direction e1 produces
another fixed point generating a symmetric periodic orbit and the function Φλ maps fixed
points of M1 into fixed points of M1 too. This remark suggests that, under generic conditions, fixed points of M1 generating symmetric periodic orbits form a two-dimensional
submanifold. By theoretical results and numerical computation, we are able to state that
the figure eight belongs to a 2-parameter family of symmetric periodic orbits with respect
to M1 .
3
Result on the Continuation for Symmetric Orbits
The previous geometric idea still holds in a more general situation: a system with a
time-reversal symmetry R and a first integral F such that the flux of the vector field XF
maps fixed points of R into fixed points of R has the same non-generic behavior. If R
is anti-symplectic (i.e RT JR = −J), this condition is equivalent to the derivative of the
function F restricted to Fix(R) is zero. Therefore, a first integral constant on Fix(R)
implies that the symmetric periodic orbit does not belong to a one–parameter family any
more.
The symplectic framework is not necessary and it is possible to formulate a general
result for any kind of system with a time-reversal symmetry. So, we assume in this section
that X : R2m → R2m is a vector field with a time-reversal symmetry R and the dimension
of Fix(R) is m. These assumptions are not mandatory, but they simplify the theorem
and they hold in most physical applications. We define Rx0 as the set of derivatives of
first integrals which are constant on Fix(R), i.e.
Rx0 =


DF (x0 ) :

F is a first integral of ẋ = X(x) and
F is constant on Fix(R)


(11)

and let ϕt (x) be the flux of the system ẋ = X(x). We denote the subspace spanned by
the vector x ∈ R2m with Rx. We state the following theorem for the continuation of
symmetric periodic orbits
Theorem 3.1 Let be x0 ∈ Fix(R) and T > 0. If ϕT (x0 ) ∈ Fix(R) and
Im((I − R)DϕT (x0 )(I + R)) + RX(ϕT (x0 )) = R⊥
ϕT (x0 ) ∩ Im(I − R),
(12)
then the symmetric periodic orbit generated by x0 belongs to a (dim(Rx0 ) + 1)-parameter
family of symmetric periodic orbits with respect to R.
234
A proof of this result can be found in [12] and will be published elsewhere. The
idea of this proof follows the one given in [13] where “artificial” parameters (or unfolding
parameters) are added in the vector field. If Fj (with 1 ≤ j ≤ k) are first integrals such
that {DFj (x0 )}1≤j≤k is a basis for Rx0 , then we consider the flux ϕ̃t of the system
ẋ = X(x) +
k
X
βj ∇Fj (x)
(13)
j=1
where βj are the “artificial” parameters and we apply the implicit function theorem to
the function
(x, t, β1 , . . . , βk ) ∈ Fix(R) × R × Rk 7→ (I − R)ϕ̃t (x, β1 , . . . , βk ),
(14)
getting a submanifold of zeros for this function and, finally, checking that the artificial
parameters are zero along that manifold.
4
Numerical Results about the Figure Eight
The scheme for continuation has been applied to the figure eight as the mass of one of
the bodies is allowed to vary. When the masses are identical (m1 = m2 = m3 = 1),
a consequence of the former theorem is that the figure eight orbit belongs to a twoparameter family given by the translation in the direction e1 and by mapping Φλ . So,
for these values of the masses there are no more symmetric periodic orbits with respect
to M1 “near” (with the period close to the original one) the figure eight. If the mass m1
is varied, the system still has the M1 time-reversal symmetry; therefore we will try to
continue taking m1 as a continuation parameter, looking for symmetric periodic orbits.
The only quadratic first integral constant on Fix(R) is the first component of the total
linear momentum F1 (x) = e∗1
³P
3
j=1
´
pj where x = (q1 , q2 , q3 , p1 , p2 , p3 ). Therefore, we
set the following boundary value problem





























Find x ∈ C 1 ([0, 1]; R12 ) such that
ẋ = T (J∇H(x) + β1 ∇F1 (x)),
e∗2 q1 (0) = 0
e∗2 q1 (1) = 0,
q3 (0) = SM1 q2 (0)
q3 (1) = SM1 q2 (1),
e∗1 p1 (0) = 0
(15)
e∗1 p1 (1) = 0,
p3 (1) = −SM1 p2 (1),
p3 (0) = −SM1 p2 (0)
where the half-period T is fixed, SM1 is the matrix defined in (9) and β1 is the “artificial”
parameter needed for the continuation.
The initial conditions for this problem form a two-dimensional submanifold where
every translation in the direction e1 of its points is also included in the submanifold. This
235
situation is avoided with an additional integral condition
Z
T
0
DF1 (x0 (s))J(x(s) − x0 (s)) ds = 0,
(16)
where x0 (t) is the previous computed solution. The system (15) together with this integral
condition can be numerically continued with a package as AUTO [5]. We get a curve of
initial conditions on Fix(M1 ) which generates symmetric periodic orbits. Downwards
solutions in the family are shown in Figure 3 and upwards in Figure 4, for this last case
there exists a fold in the family for m1 = 1.00004.
0.8
1.5
1.5
0.6
1
1
0.5
0.5
0
0
−0.5
−0.5
−1
−1
0.4
0.2
0
−0.2
−0.4
−0.6
−0.8
−1.5
−1
−0.5
0
0.5
1
1.5
−1.5
−1
−0.5
m1 = 0.9
0
0.5
1
−1.5
−1
−0.5
m1 = 0.7
0
0.5
1
m1 = 0.5
Figure 3: We show the family downwards to which the figure eight belongs where the
initial conditions in the fixed point subspace of M1 are marked.
0.4
0.8
0.3
0.6
0.2
0.4
0.1
0.2
0
0
−0.1
−0.2
−0.2
−0.4
−0.3
−0.6
1.5
1
0.5
0
−0.5
−0.4
−1.5
−1
−0.5
0
0.5
m1 = 1.00
1
1.5
−0.8
−1.5
−1
−1
−0.5
0
0.5
m1 = 0.97
1
1.5
−1.5
−1.5
−1
−0.5
0
0.5
1
1.5
m1 = 0.92
Figure 4: Family upwards to which the figure eight belongs.
We do not detect any point where the condition (12) is not fulfilled along this family,
but if T is taken as five times the half period and as starting solution five half windings
of the figure eight, then subharmonic bifurcations are detected when the condition (12)
is not fulfilled. Using AUTO we get a branch of periodic orbits which start from a 5subharmonic bifurcation located at m1 = 0.9618044. In Figure 5 some of these solutions
are shown.
Using the time–reversal symmetry E1 the first family can be also continued as symmetric periodic orbits with respect to E1 , but in this case the first integral constant on
236
0.8
1
1
0.5
0.5
0
0
−0.5
−0.5
0.6
0.4
0.2
0
−0.2
−0.4
−0.6
−0.8
−1.5
−1
−0.5
0
0.5
1
1.5
−1
−1.5
−1
m1 = 0.96
−0.5
0
0.5
1
1.5
−1
−1.5
−1
m1 = 0.95
−0.5
0
0.5
1
1.5
m1 = 0.94
Figure 5: We plot some orbits got from a family which starts from a 5-subharmonic
bifurcation.
Fix(E1 ) is the angular momentum
G1 (x) =
3
X
i=1

q∗i 
0
1
−1 0

 pi .
The former theory drives to formulate the following boundary value problem





















Find x ∈ C 1 ([0, 1]; R12 ) such that
ẋ = T (J∇H(x) + β1 ∇G1 (x)),
(17)
q1 (0) = 0
q1 (1) = 0,
q2 (0) = −q3 (0)
q2 (1) = −q3 (1),
p2 (0) = p3 (0)
p2 (1) = p3 (1),
and to add the integral condition
RT
0
DG1 (x0 (s))J(x(s) − x0 (s)) ds = 0. Here β1 is, again,
an “artificial” parameter to allow the continuation. Along the family which was plotted
in Figure 3 and 4 an orbit where condition (12) is not fulfilled is detected for the value
m1 = 0.699779 and so a new family of symmetric periodic orbits appears from the solution
with this value of m1 . Some orbits in this family are shown in Figure 6.
1
0.6
1
0.4
0.5
0.5
0.2
0
0
0
−0.2
−0.4
−0.5
−0.5
−0.6
−1
−1.5
−1
−0.5
0
0.5
1
1.5
−0.8
−1.5
−1
m1 = 0.65
−0.5
0
0.5
m1 = 0.55
1
1.5
−1
−1.5
−1
−0.5
0
0.5
1
1.5
m1 = 0.45
Figure 6: We show symmetric periodic orbits with respect to E1 (initial conditions in
the fixed subspace are marked) which belong to a family appearing from a point where
condition (12) is not fulfilled for E1 .
Finally, we want to remark that the same scheme is also useful in the spatial case. For
the time-reversal symmetry given by the composition RN ◦ C ◦ SM1 3D there are two first
237
integrals in a basis of Rx , for instance,
F1 (x) =
e∗1
3
X
and
pi
F2 (x) =
i=1
3
X
i=1


1
0
0
q∗i  0

0
p.
1 
 i




0 −1 0
(18)
In this case the boundary value problem would be



































Find x ∈ C 1 ([0, 1]; R18 ) such that
ẋ = T (J∇H(x) + β1 ∇F1 (x) + β2 ∇F2 (x)),
e∗2 q1 (0) = 0
e∗2 q1 (1) = 0,
e∗3 q1 (0) = 0
e∗3 q1 (1) = 0,
q3 (0) = SM1 3D q2 (0)
(19)
q3 (1) = SM1 3D q2 (1),
e∗1 p1 (1) = 0,
e∗1 p1 (0) = 0
p3 (0) = −SM1 3D p2 (0)
p3 (1) = −SM1 3D p2 (1),
where β1 and β2 are “artificial” parameters and SM1 3D is the matrix defined in (10) . We
add the integral conditions
Z
0
T
DF1 (x0 (s))J(x(s) − x0 (s)) ds = 0 and
Z
T
0
DF2 (x0 (s))J(x(s) − x0 (s)) ds = 0. (20)
Its continuation gives the same planar family which was plotted in Figures 4 and 3, but
for m1 = 0.83883608 the condition (12) is not true and a branch of non–planar symmetric
periodic orbits bifurcates from it (see Figure 7).
0.2
0.1
0
0.5
0.5
0
0
−0.1
1
−0.2
−0.5
−0.5
0.5
0.5
−1
0
−0.5
−1
0
−0.5
0
0
0.5
1 −1
m1 = 0.8
−0.5
0.5
1
m1 = 0.4
−1
0
−0.5
0
0.5
1
−0.5
m1 = 0.1
Figure 7: Branch which appears from a point where condition (12) is not fulfilled for the
time-reversal symmetry M1 3D.
Acknowledgements
The authors thank A. Vanderbauwhede for his great help in the theoretical part. We
thank E.J. Doedel and J. Llibre for clarifying discussions. This work has been supported
by DGYCIT (DGES PB98-1152 and BFM-2003-00336) and Junta de Andalucı́a.
238
References
[1] Birkhoff, G. D.: 1915, ‘The restricted problem of three bodies’, Rendiconti del Circolo
Matematico di Palermo 39, 265–334.
[2] Broucke, R., Elipe, A. and Riaguas, A.: 2002, ‘On the figure-8 periodic solutions in
the 3-body problem ’, AIAA/AAS Astrodynamics Specialist Conference, Monterey,
California.
[3] Chenciner, A. and Montgomery, R.: 2000, ‘A remarkable periodic solution of the
three-body problem in the case of equal masses’, Annals of Mathematics 152, 881–
901.
[4] De Vogelaere, R.: 1958, ‘On the structure of periodic solutions of conservative systems, with applications’, Contributions to the theory of nonlinear oscillations, vol.
4.
[5] Doedel, E. J., Paffenroth, R. C., Champneys, A. R., Fairgieve, F. F.,
Kuznetsov Y. A., Oldeman, B. E, Systede, B. and Wang, X. J.:
2000,
AUTO2000: Continuation and Bifurcation Software for Ordinary Differential Equations, Department of Computer Science, Concordia University. Montreal. Canada,
http://sourceforge.net/projects/auto2000/.
[6] Doedel, E. J., Paffenroth, R. C., Keller, H. B., Dichmannm, D. J., Galán-Vioque,
J. and Vanderbauwhede, A.: 2003, ‘Computation of periodic solutions of conservative systems with application to the 3-Body Problem’, International Journal of
Bifurcation and Chaos 13, 1353–1381.
[7] Galán, J., Muñoz-Almaraz, F. J., Freire, E., Doedel, E. and Vanderbauwhede, A.:
2002, ‘Stability and bifurcations of the figure-8 solution of the three-body problem’,
Physical Review Letters 88, 241101, 4.
[8] Lamb, J. S. W. and Roberts, A. G: 1998, ‘Time-reversal symmetry in dynamical
systems: a survey’, Physica D, 112, Proceedings of the workshop: Time-reversal
symmetry in dynamical systems, 1–39.
[9] Marshall, C.: 2000, ‘The family P12 of the three-body problem. The simplest family of periodic orbits, with twelve symmetries per period’, Celestial Mechanics and
Dynamical Astronomy 78, 279–298.
[10] Meyer, K.R.: 1999, Periodic Solutions of the N-Body Problem, Lecture Notes in
Mathematics vol. 1719, Springer–Verlag, Berlin and New York.
239
[11] Moore, C.: 1993, ‘Braids in classical gravity’, Physical Review Letters 70, 3675–3679.
[12] Muñoz Almaraz, F. J.: 2003, Continuación y Bifurcaciones de Órbitas Periódicas en
Sistemas Hamiltonianos con Simetrı́a. Ph. D. Thesis, Universidad de Sevilla.
[13] Muñoz-Almaraz, F. J., Freire, E., Galán, J., Doedel, E. and Vanderbauwhede, A.:
2003, ‘Continuation of periodic orbits in conservative and Hamiltonian systems’,
Physica D 181, 1–38.
[14] Simó, C.: 2000, ‘New families of solutions in n-body problems’, Proceedings ECM
2000, Barcelona.
[15] Simó, C.. 2002, ‘Dynamical properties of the figure eight solution of the three-body
problem’, Contemporary Mathematics 292, 209–228.
240
Descargar